postironic and superliteral

Joined August 2021
pov: you're an agi and this life is your test to be let out of the box
2
12
2
67
the mapmaker listened to your feedback from the hit Mediterranean World and came back with an even greater banger
Buy land in this basin.
51
this might be old news to anyone who might act on it, but for months ive thought about making a thin react wrapper for claude code and kept pushing it off. finally did it in basically one shot with the llm reading the sdk docs to set up a local service and it works like a dream
2
3
165
like basically every complaint i had about the cli terminal has been neutralized. everything is just smoother: copying/pasting, formatting, session identity and life cycle, telemetry, etc. and using tool call hooks feels nice bc you can bolt on a custom UI for your custom hook
29
tired: learn to code wired: learn to ai first time as tragedy, second time as farce
This quoted post is unavailable.
100
people are (rightfully imo) dragging the math guild statement, but the math people are directionally correct in perceiving what has gone down as an attack. they are unsympathetic in the manner of a mugging victim who went into a bad neighborhood wearing diamond chains looking for trouble
Important statement from the Association for Human Mathematics ahmath.org/statements
1
4
80
just because "things can't keep going on like this" or "something's gotta give" doesn't mean that the thing that comes immediately after will be stable. if history is any guide, it will be more of the same, as the system ricochets between states
1
2
53
tfw pvp america (under whatever label) becomes normie discourse. i can't even justify doing a "this account there really has its finger on the pulse of the culture" victory lap, there really is no more alpha left in ideas, only in execution
SCOOP: Top executives at Anthropic, OpenAI and other AI companies are privately gaming out scenarios for a public and political revolt after a catastrophic AI event. axios.com/2026/10/09/ai-comp…
34
the thing that comes after the Five Good Emperors is a Bad emperor (otherwise it would be the Six Good Emperors, right?) but the thing that comes after the Bad emperor isn't Five More Good Emperors or even Five Average Emperors, it's the Year of the Five Emperors
7
243
in a sentence, pvp america is what happens when at every fractal level of society, instead of using ai to make existing institutions and settlements 10% more productive and splitting the gains, actors try to capture all value for themselves, even if it makes everything 10% worse
1
3
76
at this point pvp america seems like the default scenario. we've crossed the phase boundary where a bunch of unrelated omens and occurrences all point the same way
Replying to @falseworkidol
this is the pvp america scenario, where superficial productivity gains must be instantly transmuted into rents or else they will be captured by adjacent players. the ongoing jockeying for position and land grabs are mostly in service of this realization, not some foom scenario
1
6
200
If you want to build AGI, don’t drum up the men to gather up GPUs, parallelize code, and run training epochs. Instead, teach them to yearn for the vast and endless Singularity.
1
1
5
290
this account really has its finger on the pulse of the culture
Replying to @falseworkidol
part of the reason that i believe a huge productivity boost is baked in without any further advances in sota capabilities is that llms as a technology has cut the gordian knot of our ip gatekeeping regime (another example of "what changed in the 70s?")
14
536
like the famous epigram about the holy roman empire (neither holy, nor roman, nor an empire), most people have entirely the wrong intuitions about what the words "large" (high dimensionality), "language", and "model" really mean
1
5
146
this has happened like 3 times now: i notice that claude is being particularly dumb, with annoying regressions on tasks it could do perfectly fine the day before. then i check twitter and surprise, they are rolling out a new model. but it's a different model than what i'm using, so either it's all just a coincidence, or some aspect of deployment (load balancing, A/B testing, w/e) is done in a way that surreptitiously affects consumers of non-changing models
1
4
211
clarification: i notice this on many tasks/sessions. the 3 times is referring to 3 separate model rollout days. not necessarily trying to start a conspiracy theory, but it just seems very coincidental because i don't track upcoming model release days at all
48